WN9 FAQ

This page contains a list of common questions about WN9. If you want a more detailed description of WN9, see the WN9 description page.

Introduction

What are the main advantages of WN9 over WN8?

• Tank-selection padding is almost eliminated.
• Formula works somewhat better for lights & arty.
• Formula doesn't blow up at the top end.
• Much smaller reroll incentive.

Why did you choose a different scale from WN8?

So that it wouldn't be confused with it. WN9 scales very differently to WN8, so the metrics could only be equal at one point: Anyone below that point would gain, and anyone above that point would lose. Also, three digits is plenty given the accuracy of these metrics.

What is account WN9 and how does it differ from overall WN8?

Account WN9 is a replacement for overall metrics that throws away each player's worst tanks (currently around 65% of battles). The goal is to reduce reroll incentive, work around problems with historical nerfs & buffs, and make it work better as a skill metric for applications that can't use recent WN9.

Does account WN9 make recent values obsolete?

No. Recent WN9 scores of 1000+ battles are much better indicators of current performance, and are useful to get an idea of recent progress.

Why does my account WN9 get a lower colour than my overall WN8?

For the average account, account WN9 and overall WN8 (using v26 expected values) are nearly equal relative to the purple mark. These are the main risk factors for dropping on the WN9 scale:

1. Rerolling (account WN9 only).
2. If good, playing a lot of medium or light tanks. If bad, playing a lot of TDs or slow heavies.
3. Playing a lot of battles in elite tier 4-6 tanks and not many in tier 10 tanks.
4. Getting very low spots per game.
5. Concentrating good performances in a handful of tanks (account WN9 only).

Why does my recent WN9 get a lower colour than my recent WN8?

That's normal. The WN9 colour scale is based on account WN9, and recent WN9 is typically closer to account WN9 than recent WN8 is to overall WN8.

Why do my per-tank WN9 values get a lower colour than my per-tank WN8 values?

Similar reason to recent values, but there's some tier variation: Many tier 10 tanks should get a similar result, but lower tiers will be 10-20% worse.

How does WN9 deal with nerfed and buffed tanks?

The expected values are based on recent data, so Recent WN9 works naturally. Account WN9 reduces the weight of any tanks that were historically nerfed, so that playing historically overpowered tanks has less long-term impact on the result.

Two per-tank WN9 values are provided for nerfed tanks: As if the tank was played recently, and as if it was played at peak historical capability. It's down to the player to know whether they played a tank when it was overpowered and interpret the results correctly.

Metric design

Why no assisted damage?

WG refuse to provide usable data. The only assisted damage parameters currently in the API are account averages, which are useless for any per-tank based results (including Account WN9) and also don't work for recent values because there's no indication of how many battles they're averaged over. It's technically possible to use assisted damage in Recent WN9 for post-8.8 accounts only, but too difficult to maintain.

Doesn't the lack of assisted damage make WN9 no better than WN8?

No. WN8 had plenty of other flaws that were worth fixing: WN9 is roughly twice as accurate as WN8 with the same data. The difference is particularly large for good players.

WG-PR includes assisted damage. Doesn't that make it better than WN9?

WG-PR has many other flaws. In practice, WG-PR has similar overall accuracy to WN8, despite widespread WN8 padding, and WN9 is far more accurate for the vast majority of players. WG-PR is slightly more accurate than WN9 for really bad players, as they have huge variations in playstyle.

Why not just use winrate?

Winrate is strongly dependent on platooning, tank selection and tier selection. It's also extremely noisy over low battle counts (typically +/-9% over 100 battles), so it's misleading for per-tank and recent data. The tank and tier selection problems can be fixed, but the other problems cannot.

I work at WG and I want to help. What do we need to add to the API to improve WN9?

All of these:

• Assisted damage per tank in some form.
• Overall battle counts over which assisted damage is recorded.
• Per-tank battle counts over which assisted damage is recorded.

It's not a problem that assisted damage is only measured over recent battles, as long as that battle count is known.

Why not cap points, capping wins games?

Adding cap points did have a small positive impact on the accuracy of the formula, but it wasn't worth the controversy or the additional code. Defence points were arguably no more useful in the formula, but they're also less controversial.

Padding

Are there penalties for playing "good" tanks?

Not if you play them well. Better tanks have higher expected values, but this should accurately track their capabilities.

Can I pad WN9 by playing MTs?

Not easily. WN9 adjusts for skill-scaling differences between tanks, so good players don't automatically pad WN9 by playing fast tanks, unlike WN8.

Can I pad WN9 by playing LTs?

Yes, but only by playstyle. LTs gain more from damage padding styles than other classes because you can sacrifice assisted damage (unmeasured) for damage (measured).

Do I get punished for playing arty?

Probably not. There are problems with the SPG data that make it less reliable than other classes. SPG skill doesn't correlate quite as well with skill in other classes, and SPG damage output dropped sharply when the "personal missions" began, probably because many arty missions encourage poor choices. Attempts were made to adjust for these problems.

Some pre-8.6 SPGs were ridiculously strong (eg. M7, SU-26, SU-5, M41), and so the difference between the two per-tank WN9 values for those SPGs can be very high.

Isn't WN9 harsh on low-tier tanks?

Well, WG is harsh on players with >2500 battles in tier 1-3 tanks, assuming that your server supports newbie matchmaking. The tier 1-3 expected values are low for sub-2500 battle players, and roughly correct for servers that don't implement newbie MM.

The tier 4-5 expected values are somewhat harsh if those tanks are played with typical starting crews, but somewhat generous for sealclubbers with skilled crews. More generally, lower-tier expected values appear high because most low-tier players are bad and/or use tanks with poor crew skills and equipment. An average tier 10 player will be far above average at tier 5.

Isn't WN9 unfair to players who don't spam premium ammo or skip stock grinds?

Less unfair than the game is. Account WN9 won't punish limited economization in the long term, as the worst-played tanks are discarded. There's no data on how much free XP a player spends or much premium ammo and consumables they use, so direct adjustments aren't possible, even if they were desirable.

Does WN9 punish platooning?

Not directly. WN9 is based primarily on damage, and to a lesser extent on frags, spots and defence. Players generally get similar results in these stats regardless of whether they platoon. However, very strong platoons (eg. triple unicum platoons in low tiers) may cannibalize each other's stats to some degree.

I'm running a clan. Which metrics should I use to filter players?

Recent WN9 can be vulnerable to smurfing unless you use thousands of battles, so you may want to use Account WN9 either in addition or as a replacement. Account WN9 requirements should be slightly lower. Other general metrics add nothing to WN9, but per-tank assisted damage may provide some additional information. You can only get that from the service record.

Of course, metrics aren't everything. You should check that recruits are playing the tanks you care about, and there are no substitutes for watching a player or playing with them.

Miscellaneous

How often are the expected values updated?

Currently once per 10 weeks, although tanks released without warning may be added sooner.

Which tanks are included in the expected value list?

Currently, any tank with battles visible in the API. This includes supertest tanks and other tanks unavailable to the general public.

What happens when I play a tank that doesn't have expected values?

• Tank WN9 cannot be generated for that tank.
• Account WN9 will ignore that tank. It won't be selected in your top 65%.
• Recent WN9 may be over or underestimated, depending on which other tanks you played recently.

Different sites have different WN9 values, why?

Recent WN9 is dependent on the interval used. WoT APIs don't provide any historical data, so sites have to sample accounts over time to generate "recent" data. Different sites won't necessarily sample accounts at the same times, and so Recent WN9 values may vary. Check the battle counts.

High-volume services such as XVM may update account and tank WN9 values infrequently. Otherwise, if two sites are generating different Account WN9 or Tank WN9 values, one is using an incorrect method or old expected values.

Where can I discuss WN9?

See the threads in the Mathematics Corner section of the WoTLabs forum

.